Geovisualization: Integration and Visualization of Multiple Datasets Using Mapbox Cecilia Cadenas California Polytechnic State University, San Luis Obispo Abstract—The use of geovisualizations to different from information visualization dealing analyze multiple datasets can ease the with abstract data spaces. [3] research and investigation to find B. Mapbox & TileMill correlations between these datasets. However, making these visualizations with The creation of these visualizations can be multiple datasets can be challenging due to challenging due to the complexity and the lack of standardization within file formats differences between the geographical data types, and the software needed to render the and how easily they can be accessed and used. visualizations. In order to help with the creation of these visualizations, many cartographers and software This project investigates the making of engineers have created many different tools that these visualizations, maps in this case, using can make the creation of these visualizations an open source platforms like Mapbox. Using easier and accessible to anyone with a computer. Python scripts for the data parsing, For the creation of the maps in this project, CartoCSS for the designing of the maps, one of these tools was used. After some research JavaScript to make the maps interactive, and and testing different tools and platforms, HTML to combine all the layers together a Mapbox was selected as the best option for the set of maps were created. These maps customization level needed for the maps wanted. contained data regarding tons of wine grape Mapbox is an online platform that makes the data crushed and yearly heat accumulation viewing of these maps simple across multiple during the growing season that demonstrate platforms since it requires no extra software to how combining multiple data sets can lead to be downloaded. Mapbox is open source and has finding important correlations between these many JavaScript libraries available to allow two separate yet related datasets. maps to be interactive. All of the needed libraries and tools needed to render a map are I. BACKGROUND AND INTRODUCTION accessible within Mapbox. Mapbox also created A. Geovisualization TileMill, a desktop application that aids with the According to the Commission on designing of maps using CartoCSS. It can help GeoVisualization of the International create the interactive maps that can then be Cartographic Association, revisualization can be uploaded to Mapbox and be accessed via the defined as the visualization of data with a web. geographical component. Often, these Because Mapbox is open source, there are geographical data have a complex structure many examples within their site that can help involving space, time, and a number of thematic with the creation of new maps. They also have attributes, which poses significant challenges to many step-by-step tutorials on how to work with the visualization. The visualization of spatial the data and how to use all their different tools data requires the use of maps or 3D displays to obtain the best possible maps. [2] where at least two display dimensions are utilized to represent the physical space, which is 1 C. Data Gathering & Parsing openheatmap. This online tool allows users to Within the last few years, obtaining data has create interactive maps in seconds by uploading become easily accessible thanks to the Internet. csv files. With this tool, maps that could change National organizations like the National by year were created fairly easy, but these maps Climatic Data Center and the National were not very customizable and multiple data Agriculture Statistics Services keep open source sets could not be added in one map. After records of the past few decades available for emailing with the creator of openheatmap about download as a csv or excel file. the issues I had with it he suggested Mapbox One of the biggest challenges when working because of the customizability it had. with any visualization tool is getting this data to Mapbox has all the tools that were needed, work with the tools. Some of the files might multiple data sets in one map, customizable, and have latitude and longitude that TileMill can interactive, but in order to take advantage of recognize. However some data is relevant to these tools, there was a learning curve to using more than just a single point on a map and is this platform. Using Mapbox requires the instead relevant to a whole region as the grape knowledge of HTML, JavaScript, and CartoCSS crush data is in this case. Also, the data in order to fully utilize all of its capabilities. obtained, might not be arranged in a way that After many tutorials and getting familiar with can be meaningful in a map if presented that the software, the creation of the maps began by way. For example, the grape crush data is making the data obtained online utilizable by obtained in yearly summaries, each grape is its Mapbox. The data gathered from the National own row and its column is a district value. To Climatic Data Center consisted of one csv file make an interesting visualization combining data containing monthly temperature averages from by grape rather than by year can be more 2000 until 2013 in stations all over California. meaningful. These averages, though important, on their own There exist a file format (JSON) that can would be difficult to visualize in a static yearly allow the creation of polygons given multiple map. To make these averages useful, the file was latitude and longitudes that can then be used to parsed and only the data during the grape make zone boundaries in maps. JSON files can growing season (April – October) were used. also store other properties within each object These values were used to compute the monthly like the average weather or the amount of grapes average heat accumulation or also commonly that were crushed in each zone. known in agriculture as growing degree values In order to make this format type and obtain that were then summed up into a yearly value for only the needed data, the original data needs to each station in California. [3,5] be parsed and modified or converted to a JSON In order to store the growing degree values file. To do this, the programming language, calculated and be able to use within TileMill, the Python was used. Python allows the scripting of latitude and longitude of each station had to be programs that can read files and create new ones converted to a JSON file format. To do this, a in the proper format needed to create useful and Google spreadsheet script was found and used informative maps. that would do take the csv file and convert it to a JSON file. [1] II. METHODOLOGY The grape data obtained from the National Because of the number of different tools out Agricultural Statistics Service did not require there to create maps, the research of a couple of any more computations in order to used, but these tools was required before creating the final since the data had been split up by wine districts rendition and obtaining all the data in the right (Figure 1), this map needed to be created in a format. Finding the right tool meant finding JSON format that TileMill could use. With the something that allowed multiple data sets to be use of an available county map in JSON a new layered, and some sort of customizable, map of the 17 districts was created by manually interactive tools for better understanding of the making the borders using the boundaries of the data. The first attempt of maps was used using counties that bordered each district. [4] 2 best weather for growing each certain type of grape. In order to make the maps interactive and allow anyone to switch between the years to see this relationship, the maps have a level of interactivity that allows the user to switch between the years as it is shown in figure 4. Users can also click on a specific district and a plot like figure 3 will show the whole time line of that district. These maps were created for each of the seven most popular wine grapes grown in California, and a relationship between each grape and its main growing location can be seen for each grape type. The grapes are: Chardonnay, Sauvignon Blanc, Cabernet Sauvignon, Merlot, Pinot Noir, Syrah, and Zinfandel. These maps can be found at Figure 1: California Grape Districts [4] http://lupo.csc.calpoly.edu/wineviz Once the district JSON maps were created, a script that parsed through the original csv files and added the grape data for that year to each object (district polygons) was developed. This script took in a minimum of four parameters, the original file grape, the year of the file, the district JSON file, and the grapes types wanted to be included. The script then creates a new JSON file with the district polygons and the grape crush data values. The new JSON data was then added to TileMill, and using CartoCSS a map layer was created for each year and each grape type. III. RESULTS As Figure 2 shows, each data set individually has meaningful information regarding the weather of California and what district crushes more tons of Chardonnay. However, when these two datasets are combined, a new set of information can be discovered that can show more meaning as to why certain areas have more Chardonnay than others. By combining these two datasets, and a timeline of 14 years, a trend can be seen and a relationship between the weather and the amount Figure 2: Heat Accumulation and Tons of of grapes that were crushed that year.